The minimum overlap problem revisited